Developer tools
faster-whisper
Paying per minute to transcribe audio that never needed to leave the machine.
Replaces
- OpenAI Whisper API$0.006/min
Prices are quoted from the vendor's own pricing page on 2026-08-23 and link to it. Vendors change pricing; check before you decide.
Install
pip install faster-whisper- source
- SYSTRAN
- licence
- MIT
- verified
- 2026-08-23
A reimplementation of OpenAI's Whisper on CTranslate2. Same model weights, run on your own hardware.
R21 transcribes meetings and client calls with this and has a standing rule never to send that audio to a paid transcription API. The cost argument is real but secondary — at $0.006 a minute, an hour of recording is about 36 cents, which nobody would notice. The reason it matters is that the recordings are client conversations, and some of them touch medical practices. Audio that stays on the machine cannot be retained, subpoenaed, or used as training data by a third party, and that is a much easier position to defend than a data-processing agreement.
Practical note: the speed claim depends on hardware. On CPU it is faster than the reference implementation but not fast. The gains people quote come from a GPU with CTranslate2's int8 quantisation.
Verified against the GitHub API on 2026-08-23: 25,046 stars, MIT.